Real-world data often follow a long-tailed distribution as the frequency of each class is typically different. For example, a dataset can have a large number of under-represented classes and a few classes with more than sufficient data. However, a model to represent the dataset is usually expected to have reasonably homogeneous performances across classes. Introducing class-balanced loss and advanced methods on data re-sampling and augmentation are among the best practices to alleviate the data imbalance problem. However, the other part of the problem about the under-represented classes will have to rely on additional knowledge to recover the missing information. In this work, we present a novel approach to address the long-tailed problem by augmenting the under-represented classes in the feature space with the features learned from the classes with ample samples. In particular, we decompose the features of each class into a class-generic component and a class-specific component using class activation maps. Novel samples of under-represented classes are then generated on the fly during training stages by fusing the class-specific features from the under-represented classes with the class-generic features from confusing classes. Our results on different datasets such as iNaturalist, ImageNet-LT, Places-LT and a long-tailed version of CIFAR have shown the state of the art performances. P. Chu et al.
Mesenchymal stem cell (MSC) administration is known to enhance the recovery of the kidney following injury. Here we tested the potential of hypoxic-preconditioned-MSC transplantation to enhance the efficacy of cell therapy on acute kidney injury (AKI) by improving MSC migration to the injured kidney. Cobalt was used as hypoxia mimetic preconditioning (HMP). MSC were subjected to HMP through 24 h culture in 200 µmol/L cobalt. Compared to normoxia cultured MSC (NP-MSC), HMP significantly increased the expression of HIF-1α and CXCR4 in MSC and enhanced the migration of MSC in vitro. This effect was lost when MSC were treated with siRNA targeting HIF-1α or CXCR4 antagonist. SPIO labeled MSC were administered to rats with I/R injury followed immediately by magnetic resonance imaging. Imaging clearly showed that HMP-MSC exhibited greater migration and a longer retention time in the ischemic kidney than NP-MSC. Histological evaluation showed more HMP-MSC in the glomerular capillaries of ischemic kidneys than in the kidneys receiving NP-MSC. Occasional tubules showed iron labeling in the HMP group, while no tubules had iron labeling in NP group, indicating the possibility of tubular transdifferentiation after HMP. These results were also confirmed by fluorescence microscopy study using CM-DiI labeling. The increased recruitment of HMP-MSC was associated with reduced kidney injury and enhanced functional recovery. This effect was also related to the increased paracrine action by HMP-MSC. Thus we suggest that by enhancing MSC migration and prolonging kidney retention, hypoxic preconditioning of MSC may be a useful approach for developing AKI cell therapy.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.